Fedora 7 to 10 Benchmarks

[MetalheadGautham]

I just discovered what might have caused the performance change from Ubuntu 7.04. I think its Gnome 2. In 7.04, gnome was 1.8 and it was very fast and responsive and only slightly heavier than Xfce (if I replaced Nautilus with PCManFM). But now its a bloat.

[daveerickson]

I just discovered what might have caused the performance change from Ubuntu 7.04. I think its Gnome 2. In 7.04, gnome was 1.8 and it was very fast and responsive and only slightly heavier than Xfce (if I replaced Nautilus with PCManFM). But now its a bloat.

Per this link:
http://www.ubuntu.com/getubuntu/releasenotes/704tour
Gnome was version 2.18.

[glasen]

What where the file systems used in the tests? Also what mount option did the have?

The tests with Ubuntu 7.04 are odd. Those packages have something different.

No the tests for 7.04 are okay. My Dell Latitude D505 is only slightly slower then the Lenovo T60 and gets roughly the same numbers for Ubuntu 8.10. But compared to the tests Michael did for Ubuntu 8.10, my D505 (P-Mobile 1.7GHz) scored 40% better than his T60, although my notebook should have higher runtimes for the tests (Singe Core vs. Dual Core, 1.7GHz vs. 1.87GHz).

So something must got terribly wrong with Michael's Notebook, because all tests after 7.04 are too slow for this kind of hardware. Just take a look in the database, search for comparable hardware and look for the numbers.

[bruno08]

Michael, can you reproduce the 7.04/10 scores on that laptop?

[tschew]

In the tests that support multithreading, what was the cpu usage? The low scores seem to crop up mostly with purely computational benchmarks.

The sequential ram read however, is inexplicable. Unless the kernel turns off dual channel in all but ubuntu 7.04 (for some weird reason). On modern intel machines this can lead to substantial performance differences in synthetic benchmarks (when mostly cached reads are used, for instance).

[bkurt78]

I think it would be quite interesting to compare the results of community based distro against a commercial one on the same hardware. Something like RHEL desktop or SLED possibly. In general one would think they would be pretty close since the commercial version is based on the community version and they are all using essentially the same kernel. Just a thought.

[airlied]

initial thoughts would be maybe SELinux and/or apparmor being the cause of some of the slowdowns.

Michael did you use the same binaries across all distros or did it get recompiled on each one... just wondering if gcc sucked more

[Sacha]

If you hadn't noticed guys, the slowdown is exactly 50% (2x). The slowdowns are only in computational benchmarks. The computer has 2 cores (read: 2x).

What could be going on? For some reason, the CPUs are only being used to half (or 1 core) of their maximum power in the latter tests.

[Michael]

initial thoughts would be maybe SELinux and/or apparmor being the cause of some of the slowdowns.

Michael did you use the same binaries across all distros or did it get recompiled on each one... just wondering if gcc sucked more

SELinux was disabled during testing.

The vast majority of the tests are built from source and rebuilt each time.

[tschew]

Hi,

I've just run the audio-encoding and memory tests on 7.04 and 8.10 and I don't see any large variations in performance other than 7.04 being 4 seconds faster in FLAC encoding. However, I've used the 1.2 test suite, so maybe there is some performance regression.